frederik
/
xserver
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

present: Fix inaccurate PresentCompleteNotify timing for TearFree 

The timing of PresentCompleteNotify events is inaccurate when a driver uses
TearFree because there's no way to know exactly when a presentation will
appear on the display without receiving a notification directly from the
driver indicating that the TearFree flip containing a presentation's pixmap
is actually visible on the display.

To fix the inaccurate PresentCompleteNotify timing, make use of the new
assumption that drivers which export TearFree permit a NULL pixmap to be
passed to their flip callback in order to make a presentation track the
exact TearFree flip responsible for rendering it onto the display.

Signed-off-by: Sultan Alsawaf <sultan@kerneltoast.com>
Acked-by: Martin Roukala <martin.roukala@mupuf.org>
This commit is contained in:
Sultan Alsawaf 2023-01-22 23:31:52 -08:00
parent d4bd39f1a5
commit 1fab978a95
1 changed files with 38 additions and 20 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

46
present/present_scmd.c
View File

@ -646,31 +646,49 @@ present_execute(present_vblank_ptr vblank, uint64_t ust, uint64_t crtc_msc)
present_execute_copy(vblank, crtc_msc); present_execute_copy(vblank, crtc_msc);
/* The presentation will be visible at the next vblank with TearFree, so /* With TearFree, there's no way to tell exactly when the presentation
* the PresentComplete notification needs to be sent at the next vblank. * will be visible except by waiting for a notification from the kernel
* If TearFree is already flipping then the presentation will be visible * driver indicating that the page flip is complete. This is because the
* at the *next* next vblank. * CRTC's MSC can change while the target MSC is calculated and even
* while the page flip IOCTL is sent to the kernel due to scheduling
* delays and/or unfortunate timing. Even worse, a page flip isn't
* actually guaranteed to be finished after one vblank; it may be
* several MSCs until a flip actually finishes depending on delays and
* load in hardware.
*
* So, to get a notification from the driver with TearFree active, the
* driver expects a present_flip() call with a NULL pixmap to indicate
* that this is a fake flip for a pixmap that's already been copied to
* the primary scanout, which will then be flipped by TearFree. TearFree
* will then send a notification once the flip containing this pixmap is
* complete.
*
* If the fake flip attempt fails, then fall back to just enqueuing a
* vblank event targeting the next MSC.
*/ */
if (!vblank->queued) { if (!vblank->queued &&
vblank->reason >= PRESENT_FLIP_REASON_DRIVER_TEARFREE) {
uint64_t completion_msc = crtc_msc + 1; uint64_t completion_msc = crtc_msc + 1;
switch (vblank->reason) { /* If TearFree is already flipping then the presentation will be
case PRESENT_FLIP_REASON_DRIVER_TEARFREE_FLIPPING: * visible at the *next* next vblank. This calculation only matters
if (vblank->exec_msc < crtc_msc) * for the vblank event fallback.
*/
if (vblank->reason == PRESENT_FLIP_REASON_DRIVER_TEARFREE_FLIPPING &&
vblank->exec_msc < crtc_msc)
completion_msc++; completion_msc++;
case PRESENT_FLIP_REASON_DRIVER_TEARFREE:
if (Success == screen_priv->queue_vblank(screen, /* Try the fake flip first and then fall back to a vblank event */
if (present_flip(vblank->crtc, vblank->event_id, 0, NULL, TRUE) ||
Success == screen_priv->queue_vblank(screen,
window, window,
vblank->crtc, vblank->crtc,
vblank->event_id, vblank->event_id,
completion_msc)) { completion_msc)) {
/* Ensure present_execute_post() runs at the next MSC */ /* Ensure present_execute_post() runs at the next execution */
vblank->exec_msc = vblank->target_msc; vblank->exec_msc = vblank->target_msc;
vblank->queued = TRUE; vblank->queued = TRUE;
} }
default:
break;
}
} }
if (vblank->queued) { if (vblank->queued) {